Review: NVIDIA GeForce RTX 2080 SUPER (Reference/Founders edition)

#5694704 Posted on: 07/27/2019 02:27 PM

---

So something I noticed that is a bit odd to me:
It only has an 8-pin and 6-pin power connector. According to the power consumption graphs, this thing is up there with the Vega 56, 780Ti, and 390X, each of which need 2x 8-pins. The 2080SF reference likely was never intended to be overclocked, but still a bit odd to me. Of course, its wattage is actually low enough to get by on what it is provided with (75W from mobo + 150W from 8-pin + 75W from 6-pin = 300W, which is sufficient headroom) but different workloads draw different amounts of power. I'm sure if you were to overclock this thing, run something like FurMark, and then somehow manage to trigger the RT cores, you'd definitely exceed the power limits.

Of course, the power limits are just simply a guideline, but, this is a reference card. I think it'd have been wise for Nvidia to expand this GPU's tolerances, and not leave room for potential worst-case scenarios.
Reference 780ti only had a 6+8 config.

#5694706 Posted on: 07/27/2019 02:39 PM

---

Ah whoops, I must have been looking at one of the non-reference cards.
So, I guess Nvidia hasn't really been doing anything all that different all these years.
Still goes to show they don't seem to want this thing OC'd.

#5694709 Posted on: 07/27/2019 02:54 PM

---

Ah whoops, I must have been looking at one of the non-reference cards.
So, I guess Nvidia hasn't really been doing anything all that different all these years.
Still goes to show they don't seem to want this thing OC'd.
Of course they don't. They don't want the RMA's involved with OCing.

#5694741 Posted on: 07/27/2019 05:55 PM

---

Ah whoops, I must have been looking at one of the non-reference cards.
So, I guess Nvidia hasn't really been doing anything all that different all these years.
Still goes to show they don't seem to want this thing OC'd.
Of course they don't. They don't want the RMA's involved with OCing.
Like what? You can use 2x 6-pin on card to draw 400W. No problem for solder, nor cables from PSU.
There are mods of PCB, that can kill current sense and allow card to draw all it wants/can at the moment.

If OC results in heat damage it will not be connector, but likely undersized VRMs with insufficient cooling. 8-pin connectors are there just to satisfy standard that's blown out of proportions.

#5694744 Posted on: 07/27/2019 06:03 PM

---

Like what? You can use 2x 6-pin on card to draw 400W. No problem for solder, nor cables from PSU.
There are mods of PCB, that can kill current sense and allow card to draw all it wants/can at the moment.

If OC results in heat damage it will not be connector, but likely undersized VRMs with insufficient cooling. 8-pin connectors are there just to satisfy standard that's blown out of proportions.
Yes, you can do that. But it's not exactly good practice for a company to create a product that doesn't conform to industry standards, even if those standards are blown out of proportion. That being said, Nvidia is actually conforming to the standards; like I said earlier, they are in fact well within the limits with their current setup. My point is just that a little bit of an OC under certain workloads and that limit will be readily exceeded.
Again, I know it'll probably be fine, and like Loophole35 said, Nvidia isn't exactly interested in or obligated to allowing people to go above the product's specs. My point was really just that, compared to similar products, this seems a bit tight for an enthusiast product.
But, I'm sure 3rd party variants will offer 2x 8-pins.

#5694749 Posted on: 07/27/2019 06:27 PM

---

Yes, you can do that. But it's not exactly good practice for a company to create a product that doesn't conform to industry standards, even if those standards are blown out of proportion. That being said, Nvidia is actually conforming to the standards; like I said earlier, they are in fact well within the limits with their current setup. My point is just that a little bit of an OC under certain workloads and that limit will be readily exceeded.
Again, I know it'll probably be fine, and like Loophole35 said, Nvidia isn't exactly interested in or obligated to allowing people to go above the product's specs. My point was really just that, compared to similar products, this seems a bit tight for an enthusiast product.
But, I'm sure 3rd party variants will offer 2x 8-pins.
Standard is simply safety overkill. 50cm wire with 1.5mm^2 area can pull 35A and V-drop on way there and back will be ~0.4V on 12V therefore something like 11.6V.
Biggest issue would be connector and potential heat at contact points. But PCIe connectors are rather good, not like old 4-pin molex.

And point remains same. Any decent PSUs have one rail for all GPUs, therefore if there is 600W available, that 8+6pin configuration is enough as long as SMDs on PCB of graphics card are willing to pull. Therefore having 8+6pin is not real obstacle from nVidia against OC. Limiting factor would be power limit in vBIOS or design of PCB itself.

#5694754 Posted on: 07/27/2019 06:43 PM

---

Standard is simply safety overkill. 50cm wire with diameter of 1.5mm can pull 35A and V-drop on way there and back will be ~0.4V on 12V therefore something like 11.6V.
Biggest issue would be connector and potential heat at contact points. But PCIe connectors are rather good, not like old 4-pin molex.
In an ideal world, sure. But not every PSU manufacturer conforms to the industry standards. The point of giving so much headroom is to account for the weakest link in the PC, or those who try to cheat the system. Not everyone uses the proper wire gauge, let alone copper wires. For multi-rail PSUs, some of them might not offer enough power over the rails the GPU has access to. Some PSUs advertise their peak wattage, but not sustained wattage. Even when you have a good PSU, you have to account for what the rest of the system handles, where vdroop might be more prominent. Of course, having an 8-pin connector doesn't guarantee these things are addressed, but it opens doors to lawsuits when it fails to comply with the standards. By using 6-pin connectors, the PSU manufacturer isn't obligated to offer any more than 75W over that connector.
All that being said, I'm not so much concerned about the GPU itself offering those last 2 pins, because I trust Nvidia knows how to make solid hardware. The problem isn't them. The overkill safety you mention only applies if the PSU is also overkill.

The whole point of giving excessive headroom is to guarantee results. Take vehicle tires for example: they can usually withstand well over 10x the pressure they advertise, but there's a reason they tell you where to stop. When you have a product that can be in such a wide range of environments, you have to account for all extremes.

Like I said, without overclocking, this GPU has more than enough headroom, so, Nvidia isn't doing anything wrong here. I actually wasn't complaining about what they were doing. I'm just saying that it doesn't leave much headroom for overclocking.

Anyway, you could argue "anybody who buys this shouldn't cheap out on a crappy PSU" but you can't control the stupidity of other people, and neither can Nvidia. By offering 2x 8-pins, you set aside all concerns. But, if Nvidia doesn't want people OC'ing this (which they probably don't) then the one 6-pin connector makes for a good deterrent (or punishment) to idiots.

#5698539 Posted on: 08/08/2019 06:23 PM

---

@Hilbert Hagedoorn hey boss, I noticed some odd data on the charts. The 2080ti numbers seem low when compared to the 2080S. The I looked back at all the old reviews of the 2080ti and noticed they were performed on an older test bench (the x99 with a 5960x at 4.2Ghz) the newest reviews are using the 9900k@default. You have a note about transitioning to the new system but are any of the benches using both systems? If so do you plan on running at least a FE 2080ti full suite again. I just don't see the 2080S being only 10% slower at QHD and the 2070S being ~17% slower as reality.